google-ai-edge/model-explorer

A modern model graph visualizer and debugger

Model Explorer – What It Is

Model Explorer is a Google‑AI‑Edge open‑source tool that lets you see inside a neural‑network model as a hierarchical graph. It parses a model file (TensorFlow, TensorFlow‑Lite, TensorFlow‑JS, MLIR, or PyTorch ExportedProgram) and draws its operations as nodes grouped into layers that you can expand or collapse. The UI runs locally (or in a Hugging Face Space) and is powered by GPU‑accelerated rendering for smooth interaction.

Core capabilities (as described in the README)

  • Hierarchical view – operations are nested in logical layers, making huge graphs easier to navigate.
  • Dynamic expand/collapse – drill down into a layer or hide details on demand.
  • Highlight I/O – quickly locate the model’s input and output nodes.
  • Metadata overlay – attach and display custom information on any node.
  • Search & duplicate‑layer detection – find nodes by name and see where identical sub‑graphs appear.
  • GPU‑accelerated rendering – smooth pan/zoom even for large models.
  • Extension framework – developers can add adapters for new model formats.
  • Command‑line & API – the tool can be launched via model-explorer or used programmatically.

Supported model formats (out‑of‑the‑box)

  • TensorFlow Lite (.tflite)
  • TensorFlow (SavedModel / GraphDef)
  • TensorFlow.js (.json + weights)
  • MLIR (Google’s intermediate representation)
  • PyTorch ExportedProgram (the Torch‑Script‑like representation used for edge deployment)

Extensibility

The project encourages community adapters. Existing community adapters include:

  • ONNX
  • VGF (Arm’s Vision Graph Format)
  • TOSA (Tensor Operator Set Architecture)

Developers can follow the Develop Adapter Extension wiki page to ship new format support.

Getting started

pip install ai-edge-model-explorer   # install the Python package
model-explorer                       # launch the UI

A web‑based version is also available on Hugging Face Spaces for quick uploads.

Where to learn more

  • Wiki – detailed installation, user guide, CLI guide, API guide, Colab notebook, and adapter development docs.
  • YouTube intro video and Google Research blog post give high‑level overviews.
  • NPM package – the visualizer component (ai-edge-model-explorer-visualizer) can be embedded in web apps.

Bottom line: Model Explorer is a genuine, production‑grade visual debugging aid for AI/ML models, aimed at engineers who need to understand, debug, or present model architectures across several popular formats.

Related

  • Project
    PINTO0309/onnx2tf

    A tool for converting ONNX models into LiteRT, TFLite, TensorFlow, and native PyTorch code to facilitate cross-framework deployment.

  • Dispatch
    Introducing HF viewer: An Interactive Visualizer for Hugging Face Models

    HF viewer is a new interactive tool designed to visualize any Hugging Face model by simply pasting its URL, offering detailed insights into model architecture at multiple granularities. This tool aims to enhance understanding and exploration of complex machine learning models for developers and researchers.

  • Project
    tensorflow/tfjs

    A hardware-accelerated JavaScript library for training and deploying machine learning models in the browser and Node.js.

  • Dispatch
    Modelmap: Interactive Visualizations of Hugging Face Model Architectures

    Modelmap is a tool that generates interactive, animated network maps of any Hugging Face model architecture without requiring the user to download model weights.

  • Project
    onnx/tensorflow-onnx

    A tool to convert TensorFlow, Keras, TensorFlow.js, and TFLite models to the ONNX format for cross-platform compatibility and deployment.